Lasers are used to anneal amorphous silicon (a-Si) to form polycrystalline silicon (p-Si). The conversion of a-Si into p-Si may be employed by heat treatment at around 1000° C. Such a procedure may only be used for a-Si on heat resistant substrates such as quartz. Such materials are expensive compared to normal float glass for display purposes. Light induced crystallization of a-Si allows the formation of p-Si from a-Si without destroying the substrate by the thermal load during crystallization. Amorphous Silicon may be deposited by a low cost process such as sputtering or chemical vapour deposition (CVD) on substrates such as glass, quartz or synthetics. The crystallization procedures are well known as excimer laser crystallization (ELC), sequential lateral solidification (SLS) or thin beam crystallization procedure (TDX). An overview of these different fabrication procedures is e.g. given by D. S. Knowles et al. “Thin Beam Crystallization Method: A New Laser Annealing Tool with Lower Cost and Higher Yield for LTPS Panels” in SID 00 Digest, 1-3; Ji-Yong Park et al. “P-60: Thin Laser Beam Crystallization method for SOP and OLED application” in SID 05 Digest, 1-3 in a brochure of the TCZ GmbH Company entitled “LCD Panel Manufacturing Moves to the next Level-Thin-Beam Directional ‘Xtallization (TDX) Improves Yield, Quality and Throughput for Processing Poly-Silicon LCDs”. Line beams with a typical size of e.g. 0.5 mm×300 mm and a homogeneous intensity distribution are for example applied in silicon annealing on large substrates using excimer lasers (ELC). State-of-the-art optical systems use refractive optical illumination systems containing crossed cylindrical lens arrays to create the desired intensity distribution. These arrays, the functionality of which is e.g. described in US 2003/0202251 A1, are examples of a more general group of homogenization schemes that divide the input beam into multiple beams using suitably shaped sub apertures. The superposition of these multiple beams in the field plane averages out intensity variations and homogenizes the beam.
Typically, two perpendicular directions are homogenized separately using cylindrical optics. The main element for each direction is a cylindrical lens array which creates a certain homogeneous angular spread. This means that for each of said direction one of said cylindrical lens arrays is optically relevant for creating said certain homogeneous angular spread. FIG. 1 shows a cross section along one of said perpendicular directions of a state-of-the-art homogenizing optical system 1 containing crossed cylindrical lens arrays. For facility reasons FIG. 1 does only show the cylindrical lens array 2 (comprising in the present case n=3 lenslets 2a, 2b, 2c) which creates said certain homogeneous angular spread (divergence) from said input beam 4 in said direction drawn (first axis direction) and indicated with an arrow 3. The range of angles 5 from each array 2 is mapped to the focal plane 6 of a subsequent condenser lens 7 to yield a homogeneous illumination 8.
The field size sf is determined by the spacing d of the first array 2 and the focal lengths fcondensor, farray of the lenses 2, 7 according to the following equation:sf=d*fcondensor/farray   (1)
The other axis of the beam is homogenized using the same technique, although other focal lengths and/or array spacings are required to obtain the desired field size.
This setup with one array 2 is sensitive to divergence (incoherence) of the incoming beam 4, which blurs the edges of the field 9, and to overcome this a second array 10 (here: n=3 lenslets 10a, 10b, 10c) with identical focal length farray can be placed in the focal plane 11 of a convex field array 2 (e.g. known from Fred M. Dickey and Scott C. Holswade, “Laser Beam Shaping”, Marcel Dekker Inc. New York/Basel 2000). This yields the classical cylindrical fly's-eye homogenizer 12 which is shown in FIG. 2.
Problem To Be Solved
In state-of-the-art systems the width of the line beam is typically some 100 times the width of a diffraction limited beam (for the given numerical aperture of the system). However, for some applications it is desirable to have a very thin (e.g., <0.05 mm) and long (e.g., >300 mm) line focus with a homogeneous intensity distribution. In this case the beam width is close to a small multiple of the diffraction limited beam size. For a Gaussian input beam (which is a good approximation for an excimer laser profile along one dimension) the diffraction limited beam size (measured at 1/e2 intensity level) iswmin=2λ/πNA   (2)wherein λ is the laser wavelength and NA the numerical aperture of the system (measured at 1/e2 intensity level of the input beam). Typical values from laser material processing are NA=0.15, and λ=308 nm giving a diffraction limited beam size wmin of approximately 1.3 μm.
Small beam widths close to this diffraction limit can not be realized with state-of-the-art homogenization using cylindrical lens arrays. A lens array divides the beam into n beams with n-times smaller width. This effectively reduces the available numerical aperture NAlenslet for each beam to 1/n of the numerical aperture NA of the system:NAlenslet=NA/n   (3)
The minimum field size that can be achieved is limited by this NAlenslet of the individual lenslets. It is well known that diffraction effects dominate the resulting beam for small beam sizes. For example for n=10 and the parameters as above, the best homogeneity of the intensity distribution one can expect from a two-stage homogenizer 12 as shown in FIG. 2 when trying to create a 10 μm field looks like the intensity distribution shown in FIG. 3 which is based on a fourier optics calculation. This rather inhomogeneous intensity distribution results from an at least almost incoherent superposition of the partial beam profiles in y-direction.
In practice, the result is further distorted by speckles due to interference between the beams from different lenslets (again known from Fred M. Dickey and Scott C. Holswade, “Laser Beam Shaping”, Marcel Dekker Inc. New York/Basel 2000). Also, two-stage homogenizers are difficult to realize for small field sizes because farray and thus the separation of the arrays need to be very large.
A homogenizer which uses a single array is easier to realize, but as stated above, it is sensitive to divergence of the incoming beam, which effectively further increases the width of the intensity distribution in the field plane. This is a problem especially for excimer lasers, where the minimum beam width that can be realized by focusing the beam is r-times the diffraction limited line width wmin, where r typically is a number between 5 and 20.
This means that with state-of-the-art solutions, very narrow intensity distributions can not be realized. It is necessary to design for a significantly larger field and use only the central part of it, (which is for example taught by Burghardt et al. in U.S. Pat. No. 5,721,416), which means that most of the light is not directed to the desired field area. The problem is to find a method and an apparatus to create a narrow, homogeneous intensity distribution with a small slope width which significantly reduces the laser power that is required.